Do not try to start a vhost supervisors on not fully booted nodes.

Sometimes when several nodes are started at the same time, add_vhost
can try to start a remote vhost supervisor on a node, which does
not have a rabbit_vhost_sup_sup process yet, resulting in `{error,rabbit_vhost_sup_sup_not_running}`
error.

Filter only fully booted nodes to start remote vhost supervisors.

2 files changed, 24 insertions, 5 deletions

diff --git a/src/rabbit.erl b/src/rabbit.erl
index 37b33bd8a4..a91f2f32b5 100644
--- a/src/rabbit.erl
+++ b/src/rabbit.erl
@@ -27,6 +27,7 @@
 -export([start/2, stop/1, prep_stop/1]).
 -export([start_apps/1, start_apps/2, stop_apps/1]).
 -export([log_locations/0, config_files/0, decrypt_config/2]). %% for testing and mgmt-agent
+-export([is_booted/0, is_booted/1, is_booting/0]).
 
 -ifdef(TEST).
 
@@ -652,10 +653,16 @@ await_startup(Node) ->
     end.
 
 is_booting(Node) ->
-    case rpc:call(Node, erlang, whereis, [rabbit_boot]) of
-        {badrpc, _} = Err -> Err;
-        undefined         -> false;
-        P when is_pid(P)  -> true
+    case rpc:call(Node, rabbit, is_booting, []) of
+        true  -> true;
+        false -> false;
+        {badrpc, _} = Err -> Err
+    end.
+
+is_booting() ->
+    case erlang:whereis(rabbit_boot) of
+        undefined        -> false;
+        P when is_pid(P) -> true
     end.
 
 wait_for_boot_to_start(Node) ->
@@ -745,6 +752,16 @@ is_running() -> is_running(node()).
 
 is_running(Node) -> rabbit_nodes:is_process_running(Node, rabbit).
 
+is_booted(Node) ->
+    case rpc:call(Node, rabbit, is_booted, []) of
+        true        -> true;
+        false       -> false;
+        {badrpc, _} -> false
+    end.
+
+is_booted() ->
+    is_running() andalso not is_booting().
+
 environment() ->
     %% The timeout value is twice that of gen_server:call/2.
     [{A, environment(A)} ||
diff --git a/src/rabbit_vhost_sup_sup.erl b/src/rabbit_vhost_sup_sup.erl
index 558648400a..993f1d8430 100644
--- a/src/rabbit_vhost_sup_sup.erl
+++ b/src/rabbit_vhost_sup_sup.erl
@@ -61,7 +61,9 @@ init([]) ->
 
 start_on_all_nodes(VHost) ->
     NodesStart = [ {Node, start_vhost(VHost, Node)}
-                   || Node <- rabbit_nodes:all_running() ],
+                   || Node <- rabbit_nodes:all_running(),
+                      %% Do not try to start a vhost on booting nodes.
+                      rabbit:is_booted(Node) ],
     Failures = lists:filter(fun
                                ({_, {ok, _}}) -> false;
                                ({_, {error, {already_started, _}}}) -> false;